View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Cancer Cell Article

Proapoptotic Function of the Retinoblastoma Tumor Suppressor

Alessandra Ianari,1,2 Tiziana Natale,2 Eliezer Calo,1 Elisabetta Ferretti,2 Edoardo Alesse,3 Isabella Screpanti,2 Kevin Haigis,4 Alberto Gulino,2,5,* and Jacqueline A. Lees1,* 1David H. Koch Institute for Integrative Cancer Research at MIT, Cambridge, MA 02139, USA 2Department of Experimental Medicine, La Sapienza University of Rome, 00161 Rome, Italy 3Department of Experimental Medicine, University of L’Aquila, 67100 L’Aquila, Italy 4Center for Cancer Research, Massachusetts General Hospital, Charlestown, MA 02129, USA 5Neuromed Institute, 86077 Pozzilli, Italy *Correspondence: [email protected] (A.G.), [email protected] (J.A.L.) DOI 10.1016/j.ccr.2009.01.026

SUMMARY

The (pRB) tumor suppressor blocks cell proliferation by repressing the transcrip- tion factors. This inhibition is relieved through mitogen-induced phosphorylation of pRB, triggering E2F release and activation of cell-cycle genes. can also activate proapoptotic genes in response to geno- toxic or oncogenic stress. However, pRB’s role in this context has not been established. Here we show that DNA damage and E1A-induced oncogenic stress promote formation of a pRB-E2F1 complex even in prolif- erating cells. Moreover, pRB is bound to proapoptotic promoters that are transcriptionally active, and pRB is required for maximal apoptotic response in vitro and in vivo. Together, these data reveal a direct role for pRB in the induction of in response to genotoxic or oncogenic stress.

INTRODUCTION In normal cells, pRB’s repressive activity is controlled by its cell-cycle-dependent phosphorylation (Trimarchi and Lees, The retinoblastoma gene (RB1), a member of the pocket protein 2002). In response to mitogenic signaling, pRB is sequentially family with p107 and p130, was the first known tumor phosphorylated by the Cdk complexes D-Cdk4/6 and suppressor. The retinoblastoma protein (pRB) is targeted by -Cdk2. This phosphorylation is sufficient to induce pRB the transforming of the DNA tumor viruses (e.g., adeno- to release E2F, thereby allowing activation of E2F-responsive viral E1A), and it is functionally inactivated in a large proportion genes in late G1. However, phosphorylated pRB (ppRB) persists of human tumor cells due to mutations of either the RB1 gene in the nucleus through the remainder of the until it is itself or its upstream regulators (Trimarchi and Lees, 2002). dephosphorylated by protein phosphatase 1 at the end of pRB’s tumor-suppressive activity is thought to be largely depen- (Ludlow et al., 1993). It is widely assumed that ppRB is dent upon its ability to directly bind members of the E2F family of functionally inactive and that dephosphorylation restores pRB transcription factors and prevent them from promoting transcrip- to the active state. The majority of human tumors carry mutations tion of genes required for cell proliferation (Trimarchi and Lees, that disable pRB-mediated repression of E2F (Sherr and McCor- 2002). This inhibition can occur via two distinct mechanisms: mick, 2002). These mutations either inactivate the RB1 gene pRB binds to sequences within E2F’s transactivation domain itself or promote pRB phosphorylation in the absence of normal and inhibits its function, and the resulting pRB-E2F complex mitogenic signals through activation of the -Cdk4/6 recruits a number of transcriptional corepressors, including kinases or inactivation of the Cdk inhibitor . These changes histone deacetylases (HDACs), methyltransferases, and poly- result in the inappropriate release of E2F, thereby inducing tran- comb group proteins to actively repress the promoters of E2F scriptional activation of E2F target genes and consequently cell target genes. proliferation.

SIGNIFICANCE

Retinoblastoma protein (pRB) function is disrupted in many human tumors through either inactivation of the RB1 gene or alterations in its upstream regulators. pRB’s tumor-suppressive activity is at least partially dependent upon its ability to arrest cells through E2F inhibition. Our data here now establish a second role for pRB as a stress-induced activator of apoptosis. Notably, pRB’s ability to promote either arrest or apoptosis seems to be context dependent, with apoptosis being favored in proliferating cells. This finding has the potential to explain why cells are typically more resistant to apoptosis when in the arrested state. Most importantly, our observations suggest that RB1 status will influence tumor response to chemotherapy by impairing both the arrest and apoptotic checkpoint responses.

184 Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc. Cancer Cell pRB Induces Apoptotic Genes in Response to Stress

It is well established that E2F1, among other E2F family that the observed DNA damage-induced formation of the pRB- members, also contributes to the induction of apoptosis in E2F1 complex is neither dependent on nor simply an indirect response to either DNA damage or oncogenic stress (Iaquinta consequence of a G1 arrest. and Lees, 2007). This is thought to be a critical event in suppress- pRB’s ability to bind to E2F is normally limited to the early ing the formation of tumors. Work from many laboratories has stages of the cell cycle when pRB exists in the hypophosphory- shown that genotoxic stress induces E2F1 recruitment to the lated form. Therefore, it was surprising to observe pRB-E2F1 promoters of proapoptotic genes including p73 and Caspase complexes in a population highly enriched for G2/M phase cells. 7, coincident with their transcriptional activation, even in cells To more directly address the influence of cell-cycle phasing, we that retain wild-type pRB (Pediconi et al., 2003). This led us to used serum deprivation and readdition to generate two popula- consider how pRB influences DNA damage-induced apoptosis. tions of T98G cells that were greatly enriched for either G0/G1 The prevailing view is that pRB is an antiapoptotic regulator. (70%) or proliferating (95% S or G2/M phase) cells (Figure 1C) Early support for this model came from the finding that several and then treated these with doxorubicin. Consistent with the tissues in Rb mutant mice display both ectopic proliferation known cell-cycle-dependent phosphorylation of pRB, the pRB and apoptosis (Jacks et al., 1992). However, it is now clear protein was present in its slower mobility form in untreated that much of this apoptosis is non-cell autonomous, resulting proliferating cells, and it bound little E2F1 (Figure 1C). Notably, from a proliferation defect in the extraembryonic tissues (de doxorubicin treatment was still able to induce formation of the Bruin et al., 2003; Wenzel et al., 2007; Wu et al., 2003). Analysis pRB-E2F1 complex in this proliferating population (Figure 1C). of tissue-specific Rb mutant models reinforces the notion that This occurred independently of any change in total levels of pRB plays a much more nuanced role in apoptosis. Loss of E2F1 protein (Figure 1C), and it correlated with full activation of pRB in neuronal tissue (MacPherson et al., 2003), lung (Mason- the apoptotic program as judged by PARP-p85 induction (data Richie et al., 2008; Wikenheiser-Brokamp, 2004), skin (Ruiz not shown). Importantly, the G0/G1 cells within the proliferating et al., 2004), and intestine (Haigis et al., 2006; Wang et al., population cannot fully account for this DNA damage-induced 2007) drives ectopic proliferation but has no effect on apoptosis. pRB-E2F1 complex because the treated proliferating cells and In contrast, Rb inactivation in the lens (de Bruin et al., 2003) and untreated G0/G1 cells had comparable levels of pRB-associated myoblasts (Huh et al., 2004) does induce apoptosis, but this is E2F1 (Figure 1C), but the fraction of G0/G1 cells in the enriched specifically observed in the differentiating cells. Thus, taken populations differed by 14-fold (5% versus 70%). Thus, these together, these mouse studies support two general conclusions: data show that pRB-E2F1 complexes can form in S/G2/M phase first, in many different settings, Rb inactivation can induce inap- cells in response to DNA damage. propriate proliferation without triggering apoptosis, and second, pRB is sequentially phosphorylated by the cyclin D-Cdk4/6 when apoptosis is observed, it seems to result from an inability to and cyclin E-Cdk2 complexes, and this is thought to disrupt cease proliferation and undergo terminal differentiation. the interaction between pRB and E2F. Since DNA damage pRB’s apoptotic role has also been analyzed in established causes pRB to bind to E2F1 irrespective of cell-cycle phase, tissue culture cell lines. However, these studies have yielded we wished to determine whether ppRB could participate in this conflicting results: some conclude that pRB suppresses complex. To this end, we generated enriched populations of apoptosis (Almasan et al., 1995; Bosco et al., 2004; Knudsen G0/G1 and proliferating T98G cells, exposed them to either et al., 2000), whereas others suggest that it is proapoptotic (Araki ionizing radiation (IR) or doxorubicin (see schema in Figure 1D, et al., 2008; Bowen et al., 1998, 2002; Knudsen et al., 1999). left panel), and then assessed both the levels and E2F1-binding None of these studies addresses the molecular basis for the properties of ppRB using antibodies that specifically recognize observed role of pRB. In this study, we investigated how pRB known Cdk phosphorylation sites within pRB, pSer780, influences the ability of E2F to induce apoptosis in response to pSer795, and pSer807–811. Fluorescence-activated cell sorting genotoxic stress. (FACS) analysis confirmed the high degree of enrichment of the G0/G1 and proliferating populations both before and after IR or RESULTS doxorubicin treatment (Figure 1D, right panel). As expected, ppRB was present at much higher levels in the proliferating Stabilization of the pRB-E2F1 Complex in Response versus the G0/G1 cells as judged by both the immunoprecipita- to DNA Damage tion (IP) of ppRB and the mobility shift of the total pRB In cells committed to die by apoptosis in response to either DNA (Figure 1E). We found that a small subset of the E2F1 coimmuno- damage or oncogenic stress, E2F1 transcriptional activity is precipitated with ppRB before treatment, and IR and doxoru- directed toward promoters of a subset of apoptotic genes that bicin both increased this level (Figure 1E, left panel). This include Caspase 7, p73, and Apaf1 (Iaquinta and Lees, 2007). increased binding occurred independently of any change in the Thus, we hypothesized that DNA damage must somehow total levels of E2F1 (Figure 1E, right panel). Interestingly, IR inactivate pRB’s repressive function to allow the release of tran- and doxorubicin increased the levels of ppRB in the G0/G1 pop- scriptionally active E2F1. To test this hypothesis, we assessed ulation (Figure 1E, compare lane 1 with lanes 2 and 3), even the binding of pRB to E2F1 both before and after doxorubicin though neither treatment altered the cell-cycle distribution of treatment of human T98G cells. However, contrary to our expec- these cells (Figure 1D, right panel). tations, we found that the pRB-E2F1 complex was stabilized The increased level of ppRB in the treated G0/G1 cells clearly by this genotoxic stress (Figure 1A). Notably, T98G cells are contributes to, but seems insufficient to fully account for, the p53 defective and therefore treatment with doxorubicin causes increased level of E2F1 in the ppRB immunoprecipitates. them to accumulate in G2/M (Figure 1B). Thus, we can conclude Indeed, for all of the cell-cycle fractions, we clearly recovered

Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc. 185 Cancer Cell pRB Induces Apoptotic Genes in Response to Stress

tail disrupts the ppRB-E2F1 complex, leading to an underesti- mation of its levels, or whether hypophosphorylated or other phosphorylated pRB species are the main constituent of this complex. To distinguish between these possibilities, we con- ducted a reciprocal IP. Specifically, we immunoprecipitated E2F1 from T98G cells before or after doxorubicin treatment and then screened for associated pRB by western blotting with an antibody that recognizes all forms of pRB (Figure 1F). In the untreated cells, the E2F1 immunoprecipitate contained a single pRB species. In contrast, doxorubicin caused E2F1 to bind two distinct pRB bands that were present at approximately equal levels. One of these comigrated with the single pRB species seen in the untreated cells, while the other had a slower mobility characteristic of ppRB. To verify this, we stripped and reprobed the blot with the anti-ppRB antibody cocktail. This recognized only the slower migrating pRB species specific to the doxorubicin-treated cells, confirming that this was ppRB. Based on the relative levels of the two bands in the anti-pRB blot, we conclude that ppRB accounts for at least half of the E2F1-associated pRB activity in the doxorubicin-treated T98G cells. Taken together, these experiments show that DNA damage induces formation of pRB-E2F1 complexes in both ar- rested and proliferating cells and that ppRB is able to participate in this complex.

pRB Participates in Transcriptional Activation of Proapoptotic Genes in Response to Stress Histone acetyltransferases (HATs) and HDACs are known to play key roles in mediating the transcriptional properties of pRB and the E2F proteins (Frolov and Dyson, 2004). The pRB-E2F complex is thought to act as a repressor of classic E2F target genes through recruitment of HDACs, while HATs acetylate E2F1 and promote its transcriptional activity. Importantly, DNA Figure 1. DNA Damage Promotes Formation of a pRB-E2F1 damage triggers the HAT P/CAF to bind and acetylate E2F1 Complex in Proliferating Cells (Ianari et al., 2004), and this modification is required for E2F1 (A and B) Asynchronous T98G cells untreated (À) or treated with 2 mM doxo- association with proapoptotic promoters (Pediconi et al., rubicin (doxo) for 24 hr (+) were screened by immunoprecipitation (IP) using 2003). Given these observations, we screened for the presence an antibody against pRB followed by western blotting (WB) to assess levels of P/CAF in pRB immunoprecipitates in untreated versus doxo- of pRB and associated E2F1 and P/CAF (A) or assayed for cell-cycle distribu- tion by FACS analysis (B). Bars in (B) represent the mean of three independent rubicin-treated cells (Figure 1A). We found that DNA damage experiments ± SD. promotes P/CAF-pRB complex formation (Figure 1A). This (C) T98G cells were highly enriched for G0/G1 or S/G2/M (prolif.) cells as deter- raised the possibility that pRB might participate in a transcrip- mined by FACS (right panel) by culturing in 0.1% FBS for 72 hr and then main- tionally active complex under proapoptotic conditions. taining in 0.1% FBS or replating in 20% FBS for 16 hr. The cells were collected To understand the transcriptional relevance of the DNA (samples 1 and 3) or treated with 2 mM doxo for additional 48 hr (samples 2 and damage-induced pRB-E2F1 complex, we examined the regula- 4) and then assayed in parallel for pRB-E2F1 complexes by IP-WB (left panel). (D and E) Enriched populations of G0/G1 or proliferating T98G cells were tion of representative cell-cycle control and proapoptotic E2F1- untreated (samples 1 and 4), irradiated (IR, 10 Gy) for 1 hr (samples 2 and 5), responsive genes in DNA-damaged T98G cells. This analysis or treated with 2 mM doxo for 24 hr (samples 3 and 6) as indicated (D, left revealed that doxorubicin caused a differential response of these panel). These samples were analyzed for cell-cycle phasing by FACS (D, right two target gene classes: activation of the proapoptotic genes panel), the presence of ppRB-E2F1 complexes by IP with either IgG control or Caspase 7 and p73 and repression of the cell-cycle regulator ppRB antibodies and then WB for pRB or associated E2F1 (E, left panel), or the (Figure 2A). To further understand this differential levels of total pRB and E2F1 of the input lysates by WB (E, right panel). response, we conducted chromatin immunoprecipitation (F) Asynchronous T98G cells, untreated (À) or treated with 2 mM doxo for 24 hr (+), were screened for the presence of ppRB-E2F1 complexes by IP (ChIP) assays. Notably, doxorubicin treatment induced pRB with either IgG control or E2F1 antibodies and then WB, first for pRB and recruitment to both the cell-cycle and proapoptotic promoters subsequently (after stripping the blot) for ppRB. A bubble in the blot yielded (Figure 2B). Quantitative analysis of these results (see the nonspecific signal (*). Figure S1 available online) showed that the increase of pRB levels was slightly higher at Caspase 7 (2-fold) and p73 (2.2- a smaller fraction of total E2F1 in ppRB immunoprecipitates fold) than at Cyclin A2 (1.5-fold) gene promoters. For all of the (Figure 1E) versus total pRB immunoprecipitates (Figure 1C). It other proteins that we assayed, doxorubicin treatment caused was unclear whether the phosphospecific pRB antibody cock- differential changes at cell-cycle versus proapoptotic promoters

186 Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc. Cancer Cell pRB Induces Apoptotic Genes in Response to Stress

HDAC1 to the proapoptotic promoters in either the damaged or undamaged cells (Figure 2B). The coordinated enrichment of pRB, E2F1, and RNA poly- merase II at the Caspase 7 and p73 promoters fits with the hypothesis that pRB contributes to activation of proapoptotic genes. However, we could not rule out the possibility that there are two distinct populations of cells in which these promoters are either bound by pRB and repressed or associated with RNA polymerase II and activated. To address this possibility, we performed ChIP-reChIP experiments in which immunopre- cipitated pRB-chromatin complexes were eluted and then sub- jected to a second round of immunoprecipitation with either control IgG or an antibody against acetyl-H4 (AcH4), a marker of transcriptional activation (Figure 2C). As with our previous experiment, the primary ChIP showed that doxorubicin promoted recruitment of pRB to both the Caspase 7 and Cyclin A2 promoters (Figure 2C). However, when we analyzed the eluate from the pRB immunoprecipitates, we found that acetyl- H4 was specifically detected at the Caspase 7, but not the Cyclin A2, gene (Figure 2C). This analysis showed unequivo- cally that pRB was bound to the transcriptionally active Caspase 7 gene promoter, presumably via its participation in the pRB- E2F1-P/CAF complex that is promoted by DNA damage. At the same time, pRB binds to cell-cycle promoters and recruits HDAC1 to mediate their repression.

pRB Is Required for Maximal Induction of the Apoptotic Response In Vitro and In Vivo These data show that pRB is associated with proapoptotic promoters that are transcriptionally active in DNA-damaged cells. However, they do not establish whether pRB contributes to tran- scriptional activation of these proapoptotic genes or to the apoptotic response. To address these questions, we took advan- tage of the lentivirus pPRIME-GFP-shRB and its control pPRIME- GFP producing respectively either a short hairpin against human pRB (shRB) or a hairpin targeting the luciferase gene, in the Figure 2. DNA Damage Induces pRB-Dependent Activation context of miR30 (Stegmeier et al., 2005; http://elledgelab. of Proapoptotic Gene Promoters (A) Untreated and doxo-treated asynchronously growing T98G cells were bwh.harvard.edu/protocols/pPRIME/pPRIME_vectors.doc). We assessed for levels of Caspase 7, p73, and Cyclin A2 mRNAs by real-time used these viruses to infect T98G cells and selected parallel pop- RT-PCR analysis. Results are normalized to GAPDH and shown relative to ulations of GFP-positive cells. The shRB reduced pRB levels to the levels observed in untreated cells (set to 1). Bars represent the mean of less than 50% of those seen in the control cells (Figure 3A). three independent experiments ± SD. Importantly, this partial knockdown had no effect on cell-cycle m (B) T98G cells were untreated (À) or treated with 2 M doxo for 16 hr (+), and phasing (Figure 3A). This allowed us to assess pRB’s contribu- the binding of pRB, E2F1, RNA polymerase II (Pol II), and HDAC1 to the Caspase 7, p73, and Cyclin A2 gene promoters was determined by ChIP. tion to apoptosis independent of its role in the cell cycle. We (C) Acetyl-H4 (AcH4) reChIP analysis of the pRB ChIP shows that pRB is found that this partial pRB knockdown significantly reduced bound to the transcriptionally active Caspase 7 gene promoter following the fraction of cells undergoing apoptosis in response to either doxo treatment. doxorubicin or another topoisomerase II inhibitor, etoposide (50% and 30% reduction, respectively; Figure 3C). This corre- (Figure 2B). At the Cyclin A2 gene promoter, we found a reduction lated with a reduction in the levels of p73 (>50%) and Caspase in the binding of both E2F1 and RNA polymerase II. In addition, 7 (20%) mRNA in the shRB-expressing cells (Figure 3D). the transcriptional corepressor HDAC1 was specifically re- Thus, we conclude that pRB loss can impair the apoptotic cruited to the Cyclin A2 promoter in treated, but not untreated, response in the absence of any cell-cycle defects. cells (Figure 2B). These changes are consistent with the All of the previous experiments were conducted in the p53- observed downregulation of Cyclin A2 mRNA and the prevailing deficient tumor cell line T98G. To determine whether our findings view that pRB mediates the transcriptional repression of cell- were more broadly relevant, we repeated this analysis in cycle promoters. At the same time, doxorubicin treatment a second tumor cell line, U2OS (Figure S2). These cells express induced the recruitment of both E2F1 (1.4- and 2.7-fold) and wild-type p53, and pRB is constitutively hyperphosphorylated RNA polymerase II (2.3- and 2-fold) to the Caspase 7 and p73 due to hypermethylation and silencing of the p16INK4a gene promoters. Importantly, we did not detect any recruitment of promoter (Park et al., 2002). In accordance with our prior results,

Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc. 187 Cancer Cell pRB Induces Apoptotic Genes in Response to Stress

Figure 3. pRB Loss Impairs the Apoptotic Response to DNA Damage (A–D) T98G cells were infected with pPRIME-GFP (GFP) or pPRIME-GFP-shRB (GFPshRB) lentivirus and sorted for >20% GFP-positive cells. (A–C) Sorted GFP and GFPshRB cells were screened for the level of pRB by WB using actin as a loading control (A), cell-cycle phasing by FACS analysis (B), or the percentage of early apoptotic cells by FACS (annexin V+, 7AADÀ) after culturing for 48 hr in either the absence (À) or pres- ence of 2 mM doxo or 25 mM etoposide (et) (C). (D) Caspase 7 and p73 mRNA levels measured by real-time RT-PCR analysis. Results are normal- ized to GAPDH and expressed relative to levels observed in GFP-infected cells (set as 1). (E) Wild-type (WT) or Rb2lox/2lox (cRb) mesenchymal stem cells (MSCs) were infected with GFP- or GFP- Cre-expressing adenoviruses. The percentage of GFP+ apoptotic cells was measured by FACS (annexin V+, 7AADÀ). Bars in (B)–(E) represent the mean of three inde- pendent experiments ± SD.

we found that numerous genotoxic agents (including doxoru- is expressed in the adult intestinal epithelium. To eliminate any bicin, etoposide, and camptothecin) triggered pRB to bind to potential contribution of Cre-mediated deletion to the DNA E2F1 and promoted both activation of proapoptotic and repres- damage response, we selected Rb+/2lox;Villin-Cre+ mice (effec- sion of cell-cycle-related E2F target genes (Figures S2A–S2C). tively Rb heterozygous) as controls for our analysis of Rb2lox/2lox; Moreover, pRB knockdown using either the pPRIME-GFP- Villin-Cre+ (Rb mutant) mice. Rb2lox/2lox;Villin-Cre+ mice are shRB lentivirus or a tTA-inducible RB hairpin impaired the known to have histologically normal intestinal crypts, but these apoptotic response and the transcriptional activation of the contain proliferating cells at ectopic locations (Kucherlapati proapoptotic gene p73 (Figures S2D–S2G). Thus, pRB can et al., 2006). Immunohistochemical staining confirmed that play a positive role in DNA damage-induced apoptosis in both pRB was expressed in the intestinal epithelium of the Rb+/2lox; the absence and the presence of p53. Villin-Cre+ controls, but not in the Rb2lox/2lox;Villin-Cre+ animals It has previously been reported that Rb inactivation renders (Figure 4A, left). We then assessed the level of proliferating cells mouse embryonic fibroblasts (MEFs) more, not less, sensitive by screening for the cell-cycle marker Ki-67 (Figure 4A, middle). to DNA damage-induced apoptosis (Almasan et al., 1995; Knud- Consistent with prior studies (Haigis et al., 2006; Kucherlapati sen et al., 2000). We have repeated these experiments in both et al., 2006), the proliferating cells were restricted to the intestinal germline and conditional Rb mutant MEFs and obtained similar crypts of the Rb+/2lox;Villin-Cre+ controls, but they existed in both results (A.I. and J.A.L., unpublished data). Thus, in different the crypts and throughout the height of the villi of Rb-deficient settings, pRB can either promote or inhibit apoptosis. It seemed intestinal epithelium. We then subjected these Rb2lox/2lox;Villin- possible that the differential consequences of pRB loss reflect Cre+ mice and their Rb+/2lox;Villin-Cre+ littermate controls to fundamental differences between tumor versus normal, or intraperitoneal (i.p.) injections of doxorubicin and examined the mouse versus human, cells. To address these possibilities, we levels of apoptosis in the proximal small intestines by staining examined a second source of primary murine cells, mesen- for cleaved caspase-3 (Figure 4A, right). As expected, apoptosis chymal stem cells (MSCs). We generated these MSCs from was essentially absent in the untreated intestinal epithelia of both mice carrying either wild-type (WT) or conditional Rb (cRb) genotypes (Figure 4A, right). In response to treatment, we alleles, infected them with adenoviruses expressing either the observed high levels of cleaved caspase-3 in the Rb+/2lox;Villin- Cre recombinase gene (+Cre) or a GFP control (ÀCre), and Cre+ control tissues. Interestingly, these apoptotic cells were confirmed recombination of the cRb alleles by PCR (data not localized exclusively within the intestinal crypts, clearly corre- shown). The four cell populations were then treated with ionizing lating with the proliferative region of the intestinal epithelium. radiation (Figure 3E) or doxorubicin (data not shown), and the This is consistent with the prevailing view that the cycling cells fraction of apoptotic cells was quantified by FACS analysis. are more predisposed to undergo apoptosis than their arrested Cre expression had no significant effect on the level of apoptosis counterparts. Notably, quantification of 60 villi from each geno- in the WT cells, but it reduced apoptosis in the cRb cells by more type showed that the fraction of proliferating cells undergoing than 30% (Figure 3E). Thus, pRB loss also impairs the apoptotic apoptosis was reduced in the Rb mutant (1.9% ± 0.88%) versus response of these primary murine cells. control (3.8% ± 0.72%) tissue. This effect was most striking in To further extend this analysis, we next examined pRB’s role in the height of the villi, where we observed few, if any, apoptotic the DNA damage response in vivo. For this, we used mice cells in the Rb2lox/2lox;Villin-Cre+ mice even though this zone carrying conditional Rb alleles and a Villin-Cre transgene, which was highly proliferative (Figure 4A). However, we did not observe

188 Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc. Cancer Cell pRB Induces Apoptotic Genes in Response to Stress

Figure 4. Conditional pRB Knockout Mice Are Less Sensitive to Genotoxic Stress (A) Analysis of proximal small intestines of Rb+/2lox;Villin-Cre+ (HET) and Rb2lox/2lox;Villin-Cre+ (MUT) mice treated with vehicle control or doxo. Left: immunostain- ing confirms Cre-mediated loss of pRb in the proximal small intestine of MUT animals. Middle panel: analysis of Ki-67, a marker of proliferating cells, shows that pRB loss causes proliferation throughout the villi. Ki-67 levels were similar in the absence and presence of doxorubicin treatment. Right: staining for cleaved caspase-3 shows high levels of apoptotic cells specifically in the base of the crypts in doxo-treated, but not untreated, WT and MUT mice. Scale bars = 50 mm in all panels. (B) The average number of apoptotic cells per intestinal crypt (±SEM) in the proximal small intestines of doxo-treated Rb+/+;Villin-Cre+ (WT; n = 5), Rb+/2lox;Villin- Cre+ (HET; n = 5), and Rb2lox/2lox;Villin-Cre+ (MUT; n = 6) mice was determined by counting cleaved caspase-3-positive cells in 21 crypts for each animal. a significant difference in the level of apoptotic cells in the E1A Promotes the Formation of Transcriptionally Active intestinal crypts of Rb2lox/2lox;Villin-Cre+ mice versus Rb+/2lox; pRB-E2F1 Complexes Villin-Cre+ controls (data not shown). Given our findings, we decided to extend our analysis to examine Since our cell studies had shown that a partial knockdown of the mechanism of action of adenoviral E1A. E1A is a potent pRB was sufficient to impair the apoptotic response, we that induces uncontrolled proliferation and also sensi- wondered whether there might be a heterozygous mutant tizes cells to apoptosis. It binds to pRB with high affinity, and this phenotype in the Rb+/2lox;Villin-Cre+ intestinal crypts. To address requires the LXCXE motif that is essential for E1A’s transforming this possibility, we subjected a second cohort of Rb+/+;Villin- activity (Helt and Galloway, 2003). The prevailing view is that E1A Cre+ (n = 5), Rb+/2lox;Villin-Cre+ (n = 5), and Rb2lox/2lox;Villin- acts to sequester pRB, allowing release of transcriptionally Cre+ (n = 6) mice to i.p. injections of doxorubicin, stained for active E2Fs. This model can explain the increased proliferation cleaved caspase-3, and quantified the level of apoptosis in the rate seen for E1A-infected cells. However, prior studies have intestinal crypts. Consistent with our prior analysis, doxorubicin shown that the interaction between E1A and pRB is insufficient treatment induced a similar apoptotic response in the intestinal to promote apoptosis in response to doxorubicin (Samuelson crypts of Rb+/2lox;Villin-Cre+ (HET) and Rb2lox/2lox;Villin-Cre+ et al., 2005). Indeed, the previous mutant analysis suggests (MUT) mice (Figure 4B). However, these two genotypes had that this requires E1A binding to both pRB and p400, a compo- a 2-fold lower level of apoptosis than the Rb+/+;Villin-Cre+ (WT) nent of the TRAAP/Tip60 HAT chromatin-remodeling complex. controls (Figure 4B). In both cases, this difference was statisti- Given our observations, we hypothesized that E1A’s proapop- cally significant (p < 0.001). Thus, in this tissue, pRB loss totic function might reflect its ability to promote formation of promotes inappropriate proliferation while reducing the ability transcriptionally active pRB-E2F1 complexes, in a manner anal- of these cells to undergo DNA damage-induced apoptosis. ogous to genotoxic stress. To test this notion, we investigated Moreover, mutation of a single Rb allele is sufficient to impair the interplay between E1A and the pRB-E2F1 complex using the apoptotic response without altering proliferation. These primary IMR-90 human diploid fibroblasts that express the observations show that Rb influences the apoptotic response murine ecotropic . We selected these cells because to DNA damage in vivo. They also raise the possibility that this they apoptose in response to genotoxic agents or E1A but are impaired response could occur in patients carrying germline more resistant to apoptosis than many other cell lines and thus RB1 mutations. are better able to tolerate E1A expression. Importantly, in

Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc. 189 Cancer Cell pRB Induces Apoptotic Genes in Response to Stress

Figure 5. pRB Facilitates E1A’s Ability to Promote Apoptosis (A) In the absence of E1A, a pRB-E2F1 complex was induced by treatment of IMR-90 cells with doxo as judged by IP for pRB and WB for pRB and E2F1. (B–I) IMR-90 cells were retrovirally transduced with either pBabe-hygro vector (ÀE1A) or pBabe-hygro-E1A 12S (+E1A), selected with 75 mg/ml hygromycin for 4 days, and then assayed as follows. (B) Cells were treated with 2 mM doxo for 16 hr, and WB was used to determine the levels of pRB, E2F1, E2F3A, , and E1A in pRB IPs (left) and total lysates (right). A longer exposure of the key input lanes is shown at far right. E1A expres- sion strongly increased both the total levels of pRB, E2F1, and E2F3A and the level of pRB- E2F1/E2F3A complexes. At the exposure shown, the pRB-E2F1 interaction is not visible in the ÀE1A +doxo cells. (C) The binding of E1A to pRB and E2F1 is also detected via IP of E1A. (D) IPs with antibodies against ppRB and WB for pRB and E2F1. (E) The levels of pRB, E1A, E2F1, and AcH4 asso- ciated with the p73 and Cyclin A2 gene promoters were determined by ChIP. Densitometric quantifi- cation of each signal relative to the input is shown. (F) Real-time RT-PCR analysis of p73 and Cyclin A2 mRNA levels. Results are expressed as arbi- trary units normalized to GAPDH and show the mean of three independent experiments ± SD. (G–I) Control or E1A-transduced cells were in- fected with either GFP or GFPshRB lentiviruses. (G) WB confirmed a reduction in pRB levels after shRB infection, using actin as a loading control. (H) Caspase 7, p73, and Cyclin A2 mRNA levels were determined by real-time RT-PCR analysis in cells cultured in the absence (À) or presence of 2 mM doxo for 12 hr (+). Results are normalized to GAPDH and show the mean of three independent experiments ± SD, expressed relative to levels observed in the untreated cells (set as 1). (I) FACS analysis of early apoptotic cells (annexin V+, 7AADÀ) after culture in the absence (À) or presence of 1 mM etoposide (et) for 16 hr. Values represent the percentage of apoptotic GFP+ cells and represent the mean of three independent experiments ± SD. a manner similar to T98G and U2OS cells, doxorubicin promotes to our analysis of the DNA damage response, E1A also enhanced formation of pRB-E2F1 complexes in IMR-90 cells in the the formation of ppRB species and their binding to E2F1 absence of E1A (Figure 5A). To study the role of E1A, IMR-90 (Figure 5D). Additionally, E1A yielded a dramatic increase in cells were retrovirally transduced with either pBabe-puro vector the intracellular levels of pRB, E2F1, and E2F3A, but not E2F4 or pBabe-puro expressing the 12S form of E1A. We found that (Figure 5B). DNA binding is known to protect E2F-pocket protein E1A and E2F1 both coimmunoprecipitated with pRB complexes from degradation (Hofmann et al., 1996). Given the (Figure 5B). Doxorubicin treatment caused a modest additional observed stabilization of both pRB and the activating E2Fs in increase (1.6-fold) in the level of E2F1 coimmunoprecipitating E1A-expressing cells, we hypothesized that E1A induces the with pRB in the E1A-expressing IMR-90 cells (Figure 5B). These formation of pRB-E2F1-E1A complexes that bind to DNA and observations show that E1A potently induces formation of pRB- activate transcription. Thus, we used ChIP assays to assess E2F1-E1A complexes and that exogenous DNA damage binding to the p73 and Cyclin A2 promoters (Figure 5E). In the reinforces this response. absence of E1A, we observed a significant pRB ChIP signal at In the presence of E1A, pRB also associated with E2F3A, both promoters, but little or no binding of either E2F1 or acetyl- another activating E2F that is known to participate in the onco- H4 (Figure 5E). Since the uninfected IMR-90 cells are predomi- genic stress response, but not with the repressive E2F E2F4 nantly in G0/, we speculate that pRB contributes to (Figure 5B). Importantly, the reciprocal immunoprecipitation the repression of both p73 and Cyclin A2 in this setting. Accord- using antibodies against E1A also recovered both pRB and ingly, these genes appeared to be transcriptionally silent, as E2F1 (Figure 5C). Since it is well established that a pocket judged by the lack of p73 and Cyclin A2 mRNA (Figure 5F). In protein, such as pRB, is required to bridge the interaction contrast, in the E1A-infected cells, pRB, E2F1, E1A, and between E1A and E2F (Fattaey et al., 1993), we can infer that acetyl-H4 all associated with both the Cyclin A2 and p73 these three proteins must be part of the same complex. Similar promoters, coincident with the dramatic induction of both of

190 Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc. Cancer Cell pRB Induces Apoptotic Genes in Response to Stress

Figure 6. Model of pRB-E2F1 Complexes Involved in the Regulation of Proliferation and Proapoptotic Genes in Response to DNA Damage and E1A-Induced Oncogenic Stress See text for details.

knockdown and genetic ablation experi- ments confirm that pRB loss typically reduces the apoptotic response to DNA damage by 34%–50%. Since the apoptotic threshold is determined by these mRNAs (Figures 5E and 5F). Given the existence of a pRB- many different factors, the degree of impairment is striking. E2F1-E1A complex and the positive ChIP signals for E1A, pRB, Moreover, this is observed in many different settings including and E2F1, we speculate that E1A’s ability to activate apoptosis both primary and tumor cells derived from either mice or hu- and cell-cycle genes reflects, at least in part, its direct action mans, and it is independent of p53. Thus, our finding that pRB at their promoters. has proapoptotic activity has broad relevance. As noted above, the current view of E1A action is that it acts to The concept that pRB can participate in transcriptionally sequester pRB and thereby release transcriptionally active E2F. active complexes is not without precedent, since pRB has If this model fully explains the relationship between E1A and been shown to cooperate with differentiation-specific transcrip- pRB, E1A’s ability to induce apoptosis should not be impaired tion factors in the activation of key target genes (Charles et al., by pRB loss. To test this, we performed a pRB knockdown in 2001; Gery et al., 2004; Thomas et al., 2001). However, this proa- E1A-infected IMR-90 cells. The shRB yielded significant, but poptotic function of pRB is not observed in all situations. As we not complete, pRB knockdown in both uninfected and E1A- outlined in the introduction, the analysis of Rb mutant embryos infected IMR-90 cells (Figure 5G). Strikingly, the shRB caused led to the prevailing view that pRB is antiapoptotic. Although a 2-fold reduction in the levels of Caspase 7 and p73 mRNAs much of this apoptosis is non-cell autonomous (de Bruin et al., in either the absence or presence of DNA damage (Figure 5H). 2003; Wenzel et al., 2007; Wu et al., 2003), pRB loss does Consistent with previous studies (Samuelson et al., 2005), promote apoptosis in a cell-autonomous manner in some tissues expression of E1A alone induced only low levels of apoptosis of the developing embryo, and this seems to reflect an inability to in these cells, and this was unaffected by pRB levels undergo terminal differentiation (de Bruin et al., 2003; Haigis (Figure 5I). However, when combined with genotoxic stress, et al., 2006; Huh et al., 2004; MacPherson et al., 2003; Mason- E1A induced programmed cell death at 2-fold higher levels in Richie et al., 2008; Ruiz et al., 2004; Wang et al., 2007; Wikenhe- control versus shRB-expressing cells (Figure 5I). Thus, pRB iser-Brokamp, 2004). Moreover, some cell types, such as plays a positive role in E1A-induced apoptosis. Taken together, MEFs, have a heightened sensitivity to DNA damage in the our findings suggest that E1A associates with both pRB and absence of pRB. Thus, taken together, the existing literature E2F1, stabilizing these proteins, and the resulting complex asso- and the present study indicate that pRB can either suppress or ciates with both proapoptotic promoters to promote their tran- promote apoptosis, depending on the cellular context. We scription. note that there is a strong correlation between the proliferative properties of the cell and the observed role of pRB in apoptosis. DISCUSSION Thus, we propose the following model of pRB action (Figure 6). In G0/G1 cells, genotoxic stress induces pRB recruitment into the The E2F transcription factors play a key role in promoting cellular classic repressive pRB-E2F-HDAC complex. This prevents proliferation under the control of the pRB tumor suppressor. It is cell-cycle entry and thus acts indirectly to protect cells from well established that E2F1, among other E2F family members, apoptosis. Thus, in G0/G1 cells, pRB loss would impair arrest also contributes to the induction of apoptosis in response to and thereby promote apoptosis. In contrast, in proliferating cells, either DNA damage or oncogenic stress (Iaquinta and Lees, genotoxic stress favors formation of the transcriptionally active 2007). However, the role of pRB in this process is poorly under- pRB-E2F1-P/CAF complexes because the hyperphosphoryla- stood. We anticipated that DNA damage would have to release tion of pRB inhibits its participation in the repressive complexes. E2F1 from pRB to allow it to activate proapoptotic genes. Consequently, in this setting, pRB is proapoptotic. Interestingly, Instead, our data suggest an unexpected mechanism of pRB this context-dependent model of pRB function has the potential action—that DNA damage induces pRB to participate in to explain the well-established phenomenon that proliferating a transcriptionally active complex that drives expression of cells have a greater predisposition to undergo apoptosis proapoptotic genes. This model is supported by three central compared to their quiescent counterparts. Importantly, both observations: (1) pRB is induced to bind both E2F1 and the our cell line and in vivo studies show that a reduction in pRB histone acetylase P/CAF in DNA-damaged cells; (2) ChIP-reChIP levels is sufficient to impair pRB’s proapoptotic function without assays show unequivocally that pRB is bound to the promoters any disruption of proliferation control. This was particularly of proapoptotic genes that are transcriptionally active; and (3) striking in our animal experiments, where Rb haploinsufficiency

Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc. 191 Cancer Cell pRB Induces Apoptotic Genes in Response to Stress

reduced the apoptotic response to doxorubicin as efficiently as apoptosis and cell-cycle-related genes and to induce apoptosis the complete inactivation of Rb. This dose-dependent effect in response to DNA-damaging agents. raises the possibility that mutation of a single Rb allele might The elucidation of the mechanism by which pRB acts as increase the probability of cellular transformation by impairing a tumor suppressor has been complicated by various factors. the apoptotic response to DNA damage and thereby enabling In addition to its role in cell-cycle control, pRB has been impli- the acquisition of mutations within other genes. If this is true, cated in regulating a wide variety of cellular processes, including individuals carrying germline RB1 mutations would be particu- DNA replication, differentiation, and apoptosis (Classon and larly at risk. Harlow, 2002). Whereas the decreased differentiation potential Clearly, additional questions remain about how genotoxic and the increase in proliferative rate observed in pRB-deficient stress triggers formation of the pRB-E2F1-P/CAF complex and cells could contribute to tumorigenesis, it is more difficult to directs it specifically to proapoptotic promoters. Our data reconcile pRB’s role as a tumor suppressor with the notion show that ppRB can participate in the DNA damage-induced that loss of pRB may lead to increased apoptosis. Our finding pRB-E2F1 complexes. However, it is still unclear whether pRB that pRB plays a positive role in DNA damage-induced apoptosis phosphorylation actively promotes or is merely permissive for widens our understanding of pRB’s functions. Notably, the formation of the proapoptotic pRB-E2F1-P/CAF complex. More- behavior of pRB in the DNA damage response bears strong over, although we conducted these studies using antibodies parallels to that of p53: both of these tumor suppressors appear against known Cdk phosphorylation sites (Ser780, Ser795, and capable of triggering either cell arrest or apoptosis, depending Ser807–811), we cannot be sure whether cyclin/Cdk complexes on the cellular context. Given this model, we propose that RB1 or other kinases are responsible for this modification in the DNA- inactivation in tumor cells promotes tumorigenicity by yielding damaged cells. Furthermore, it is entirely possible that additional both a proliferative advantage and resistance to apoptotic stimuli posttranslational modifications of pRB and/or E2F1 may facili- such as chemotherapeutic treatments. tate formation of the proapoptotic complex. We note that pRB has been shown to contain a second E2F binding site that EXPERIMENTAL PROCEDURES does not interfere with E2F1’s transactivation domain (Dick and Dyson, 2003). Thus, it is intriguing to speculate that DNA Cell Culture and Infections damage somehow induces pRB and E2F1 to adopt this alternate T98G, U2OS, and IMR-90 cell lines were cultured in DMEM with 10% heat-in- structure. If this model is true, this conformation must also activated FBS. The IMR-90 cells overexpressed the murine ecotropic enable recruitment of P/CAF, which is known to be critical for receptor. Lentiviral and retroviral preparations and infections were performed E2F1-dependent activation of proapoptotic target genes in as described previously (Samuelson et al., 2005; Stegmeier et al., 2005). Mesenchymal stem cells were generated by mechanically crushing femurs response to DNA damage (Ianari et al., 2004; Pediconi et al., and tibias of 6- to 8-week-old mice and culturing in a-MEM with 10% heat- 2003). inactivated FBS. These cells were infected with either Ad5CMVCre-eGFP or Our data also cause us to revise our view of E1A’s mechanism Ad5CMVeGFP at about 100 plaque-forming units per cell for 4 hr (University of action. The prevailing view holds that E1A acts to disrupt pRB- of Iowa Gene Transfer Vector Core) and treated 3 days later with 2 mM doxo- E2F complexes, releasing free E2F1 to induce transcription of its rubicin or irradiated for 15 min for a total dose of 10 Gy and analyzed by FACS target genes. However, our data show that E1A forms a stable after 24 hr. Hairpins used in this study are shown in Table S1. complex with both pRB and the activating E2Fs. Moreover, pRB, E2F1, and E1A are all recruited to the promoters of both FACS Analysis apoptosis and cell-cycle genes, coincident with their transcrip- Suspensions of T98G or IMR90 cells were processed for DNA content as tional activation. Interestingly, mutant analysis has shown that described previously (Pozarowski and Darzynkiewicz, 2004). For apoptosis E1A must interact with both pRB and p400 in order to promote assays, cell suspensions were stained with annexin V APC and 7AAD (Becton Dickinson). Cells were analyzed using a FACScan system (Becton Dickinson), apoptosis in response to doxorubicin (Samuelson et al., 2005). and the data were analyzed using ModFit LT software (Verity Software). Thus, we now propose that E1A’s ability to promote both apoptosis and proliferation reflects at least in part its direct action at proapoptotic and cell-cycle gene promoters through Immunoprecipitations and Western Blotting its association with the pRB-E2F1 complex and the concomitant Proteins were extracted with RIPA buffer (Pediconi et al., 2003) and quantified using BCA protein assay reagent (Pierce). Extracts were immunoprecipitated recruitment of p400 and other transcriptional coactivators with the indicated antibodies and either protein A or protein G Plus (Santa Cruz (Figure 6). Given the recent finding that oncogenic stress acti- Biotechnology). Antibodies were obtained from Santa Cruz Biotechnology vates the DNA damage response (Bartkova et al., 2006; Di Micco (E2F1 [C-20], RNAPol-II [N-20], E1A [M73-HRP], actin [I-19], [C-18], et al., 2006), formation of the transcriptionally active pRB-E2F1 and E2F4 [C-20]), Cell Signaling Technology (RB [4-H1], 780-795-807-811 complexes may be further reinforced through E1A’s activation ppRB, and cleaved caspase-3), BD Pharmingen (pRB and Ki-67), Upstate of the DNA damage-dependent process. Essentially, E1A would Biotechnology (acetyl-H4 and HDAC1), and P. Nakatani of Dana-Farber Cancer Institute, Harvard Medical School (P/CAF) . trigger the DNA damage response and then piggyback onto the resultant pRB-E2F1 complex to create a stable superactivator. Notably, in contrast to the DNA damage-induced pRB-E2F1 ChIP Assay complex, which specifically activates only proapoptotic genes, Chromatin immunoprecipitation was performed as described previously (Pediconi et al., 2003). For reChIP experiments, RB immunoprecipitates the E1A-containing species has an expanded target specificity were eluted with DTT and then subjected to a second round of immunoprecip- that now includes both apoptosis and cell-cycle targets. Impor- itation with acetyl-H4 antibody or with IgG. Densitometric quantification of tantly, in agreement with this superactivator model, pRB knock- ChIP results was performed using the NIH ImageJ 1.4 program. Primer down impairs E1A’s ability to promote the transcription of both sequences are described in Table S2.

192 Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc. Cancer Cell pRB Induces Apoptotic Genes in Response to Stress

Real-Time RT PCR Bowen, C., Birrer, M., and Gelmann, E.P. (2002). Retinoblastoma protein-medi- Total RNA was extracted with a QIAGEN RNeasy Kit and reverse transcribed ated apoptosis after gamma-irradiation. J. Biol. Chem. 277, 44969–44979. with oligo(dT) primers and SuperScript II reverse transcriptase (Invitrogen). Charles, A., Tang, X., Crouch, E., Brody, J.S., and Xiao, Z.X. (2001). Retino- Quantitative analysis of Caspase 7, TAp73, and Cyclin A2 mRNA expression blastoma protein complexes with C/EBP proteins and activates C/EBP-medi- was performed employing an ABI PRISM 7900 Sequence Detection System ated transcription. J. Cell. Biochem. 83, 414–425. (Applied Biosystems). values were normalized to GAPDH. Results are expressed as mean ± SD. Statistical differences were analyzed Classon, M., and Harlow, E. (2002). The retinoblastoma tumour suppressor in by Mann-Whitney nonparametric test. p < 0.05 was considered statistically development and cancer. Nat. Rev. Cancer 2, 910–917. significant. Danielian, P.S., Bender Kim, C.F., Caron, A.M., Vasile, E., Bronson, R.T., and Lees, J.A. (2007). E2f4 is required for normal development of the airway epithe- Animal Maintenance and Tissue Analyses lium. Dev. Biol. 305, 564–576. 2lox + The Rb ;Villin-Cre mice were maintained and genotyped as described de Bruin,A., Wu,L., Saavedra, H.I., Wilson,P.,Yang, Y.,Rosol,T.J., Weinstein,M., previously (Haigis et al., 2006). The relevant genotypes were injected intraper- Robinson, M.L., and Leone, G. (2003). Rb function in extraembryonic lineages itoneally with either saline vehicle (0.9% NaCl) or doxorubicin (10 mg/kg) and suppresses apoptosis in the CNS of Rb-deficient mice. Proc. Natl. Acad. Sci. sacrificed 3 hr later, and intestines were collected for histology. All animal USA 100, 6546–6551. procedures followed protocols approved by MIT’s Committee on Animal Dick, F.A., and Dyson, N. (2003). pRB contains an E2F1-specific binding Care. pRB immunostaining was conducted using an UltraVision LP Detection domain that allows E2F1-induced apoptosis to be regulated separately from System (Lab Vision Corporation) with the primary antibody (G3-245, BD) at other E2F activities. Mol. Cell 12, 639–649. a concentration of 1:100. Immunohistochemistry was performed as described previously for cleaved caspase-3 (Haigis et al., 2006) and Ki-67 (Danielian Di Micco, R., Fumagalli, M., Cicalese, A., Piccinin, S., Gasparini, P., Luise, C., et al., 2007). All samples were counterstained with hematoxylin. Schurra, C., Garre, M., Nuciforo, P.G., Bensimon, A., et al. (2006). Oncogene- induced senescence is a DNA damage response triggered by DNA hyper- replication. Nature 444, 638–642. SUPPLEMENTAL DATA Fattaey, A.R., Harlow, E., and Helin, K. (1993). Independent regions of adeno- The Supplemental Data include two tables and two figures and can be virus E1A are required for binding to and dissociation of E2F-protein found with this article online at http://www.cancercell.org/supplemental/ complexes. Mol. Cell. Biol. 13, 7267–7277. S1535-6108(09)00033-6. Frolov, M.V., and Dyson, N.J. (2004). Molecular mechanisms of E2F-depen- dent activation and pRB-mediated repression. J. Cell Sci. 117, 2173–2181. ACKNOWLEDGMENTS Gery, S., Gombart, A.F., Fung, Y.K., and Koeffler, H.P. (2004). C/EBPepsilon interacts with retinoblastoma and E2F1 during granulopoiesis. Blood 103, Vectors and cell lines were kindly provided by F. Stegmeier (pPRIME-CMV- 828–835. GFP and pPRIME-CMV-GFP-shRB) and S. Lowe (pBabe-hygro, pBabe- E1A-hygro, and U2OS rtTA RB.670). We thank K. Hilgendorf and A. Cheung Haigis, K., Sage, J., Glickman, J., Shafer, S., and Jacks, T. (2006). The related for technical help and M. Levrero, Koch Institute colleagues, and members retinoblastoma (pRb) and p130 proteins cooperate to regulate homeostasis in of the Gulino and Lees laboratories for helpful discussions throughout this the intestinal epithelium. J. Biol. Chem. 281, 638–647. work. This project was supported by grants from the Associazione Italiana Helt, A.M., and Galloway, D.A. (2003). Mechanisms by which DNA tumor per la Ricerca sul Cancro; Telethon (GGP07118); Istituto Pasteur Cenci Bolog- virus oncoproteins target the Rb family of pocket proteins. Carcinogenesis netti; Ministero dell’Istruzione, Universita` e della Ricerca; the Italian Ministry of 24, 159–169. Health; and the Rome Oncogenomic Center to A.G. and from the NIH (2-P01- Hofmann, F., Martelli, F., Livingston, D.M., and Wang, Z. (1996). The retino- CA42063) to J.A.L. A.I. was supported by postdoctoral fellowships from the blastoma gene product protects E2F-1 from degradation by the ubiquitin-pro- American-Italian Cancer Foundation and from the Marie Curie Outgoing Inter- teasome pathway. Genes Dev. 10, 2949–2959. national Fellowship. J.A.L. is a Ludwig Scholar at MIT. Huh, M.S., Parker, M.H., Scime, A., Parks, R., and Rudnicki, M.A. (2004). Rb is Received: March 20, 2008 required for progression through myogenic differentiation but not maintenance Revised: August 3, 2008 of terminal differentiation. J. Cell Biol. 166, 865–876. Accepted: January 26, 2009 Ianari, A., Gallo, R., Palma, M., Alesse, E., and Gulino, A. (2004). Specific role Published: March 2, 2009 for p300/CREB-binding protein-associated factor activity in E2F1 stabilization in response to DNA damage. J. Biol. Chem. 279, 30830–30835. REFERENCES Iaquinta, P.J., and Lees, J.A. (2007). Life and death decisions by the E2F tran- scription factors. Curr. Opin. Cell Biol. 19, 649–657. Almasan, A., Yin, Y., Kelly, R.E., Lee, E.Y., Bradley, A., Li, W., Bertino, J.R., and Jacks, T., Fazeli, A., Schmitt, E.M., Bronson, R.T., Goodell, M.A., and Wahl, G.M. (1995). Deficiency of retinoblastoma protein leads to inappropriate Weinberg, R.A. (1992). Effects of an Rb mutation in the mouse. Nature 359, S-phase entry, activation of E2F-responsive genes, and apoptosis. Proc. Natl. 295–300. Acad. Sci. USA 92, 5436–5440. Araki, K., Ahmad, S.M., Li, G., Bray, D.A., Jr., Saito, K., Wang, D., Wirtz, U., Knudsen, K.E., Weber, E., Arden, K.C., Cavenee, W.K., Feramisco, J.R., and Sreedharan, S., O’Malley, B.W., Jr., and Li, D. (2008). Retinoblastoma RB94 Knudsen, E.S. (1999). The retinoblastoma tumor suppressor inhibits cellular enhances radiation treatment of head and neck squamous cell carcinoma. proliferation through two distinct mechanisms: inhibition of cell cycle progres- Clin. Cancer Res. 14, 3514–3519. sion and induction of cell death. Oncogene 18, 5239–5245. Bartkova, J., Rezaei, N., Liontos, M., Karakaidos, P., Kletsas, D., Issaeva, N., Knudsen, K.E., Booth, D., Naderi, S., Sever-Chroneos, Z., Fribourg, A.F., Vassiliou, L.V., Kolettas, E., Niforou, K., Zoumpourlis, V.C., et al. (2006). Onco- Hunton, I.C., Feramisco, J.R., Wang, J.Y., and Knudsen, E.S. (2000). RB- gene-induced senescence is part of the tumorigenesis barrier imposed by dependent S-phase response to DNA damage. Mol. Cell. Biol. 20, 7751–7763. DNA damage checkpoints. Nature 444, 633–637. Kucherlapati, M.H., Nguyen, A.A., Bronson, R.T., and Kucherlapati, R.S. Bosco, E.E., Mayhew, C.N.,Hennigan, R.F., Sage,J., Jacks, T., andKnudsen,E.S. (2006). Inactivation of conditional Rb by Villin-Cre leads to aggressive tumors (2004). RB signaling prevents replication-dependent DNA double-strand breaks outside the gastrointestinal tract. Cancer Res. 66, 3576–3583. following genotoxic insult. Nucleic Acids Res. 32, 25–34. Ludlow, J.W., Glendening, C.L., Livingston, D.M., and DeCarprio, J.A. (1993). Bowen, C., Spiegel, S., and Gelmann, E.P. (1998). Radiation-induced Specific enzymatic dephosphorylation of the retinoblastoma protein. Mol. Cell. apoptosis mediated by retinoblastoma protein. Cancer Res. 58, 3275–3281. Biol. 13, 367–372.

Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc. 193 Cancer Cell pRB Induces Apoptotic Genes in Response to Stress

MacPherson, D., Sage, J., Crowley, D., Trumpp, A., Bronson, R.T., and Jacks, T. Stegmeier, F., Hu, G., Rickles, R.J., Hannon, G.J., and Elledge, S.J. (2005). A (2003). Conditional mutation of Rb causes cell cycle defects without apoptosis lentiviral microRNA-based system for single-copy polymerase II-regulated in the central nervous system. Mol. Cell. Biol. 23, 1044–1053. RNA interference in mammalian cells. Proc. Natl. Acad. Sci. USA 102, Mason-Richie, N.A., Mistry, M.J., Gettler, C.A., Elayyadi, A., and Wikenheiser- 13212–13217. Brokamp, K.A. (2008). Retinoblastoma function is essential for establishing Thomas, D.M., Carty, S.A., Piscopo, D.M., Lee, J.S., Wang, W.F., Forrester, lung epithelial quiescence after injury. Cancer Res. 68, 4068–4076. W.C., and Hinds, P.W. (2001). The retinoblastoma protein acts as a transcrip- Park, Y.B., Park, M.J., Kimura, K., Shimizu, K., Lee, S.H., and Yokota, J. (2002). tional coactivator required for osteogenic differentiation. Mol. Cell 8, 303–316. Alterations in the INK4a/ARF locus and their effects on the growth of human osteosarcoma cell lines. Cancer Genet. Cytogenet. 133, 105–111. Trimarchi, J.M., and Lees, J.A. (2002). Sibling rivalry in the E2F family. Nat. Rev. Mol. Cell Biol. 3, 11–20. Pediconi, N., Ianari, A., Costanzo, A., Belloni, L., Gallo, R., Cimino, L., Porcellini, A., Screpanti, I., Balsano, C., Alesse, E., et al. (2003). Differential Wang, Y., Ray, S.K., Hinds, P.W., and Leiter, A.B. (2007). The retinoblastoma regulation of E2F1 apoptotic target genes in response to DNA damage. Nat. protein, RB, is required for gastrointestinal endocrine cells to exit the cell cycle, Cell Biol. 5, 552–558. but not for hormone expression. Dev. Biol. 311, 478–486. Pozarowski, P., and Darzynkiewicz, Z. (2004). Analysis of cell cycle by flow Wenzel, P.L., Wu, L., de Bruin, A., Chong, J.L., Chen, W.Y., Dureska, G., Sites, cytometry. Methods Mol. Biol. 281, 301–311. E., Pan, T., Sharma, A., Huang, K., et al. (2007). Rb is critical in a mammalian Ruiz, S., Santos, M., Segrelles, C., Leis, H., Jorcano, J.L., Berns, A., tissue stem cell population. Genes Dev. 21, 85–97. Paramio, J.M., and Vooijs, M. (2004). Unique and overlapping functions of pRb and p107 in the control of proliferation and differentiation in epidermis. Wikenheiser-Brokamp, K.A. (2004). Rb family proteins differentially regulate Development 131, 2737–2748. distinct cell lineages during epithelial development. Development 131, 4299– Samuelson, A.V., Narita, M., Chan, H.M., Jin, J., de Stanchina, E., McCurrach, M.E., 4310. Fuchs, M., Livingston, D.M., and Lowe, S.W. (2005). p400 is required for E1A to Wu, L., de Bruin, A., Saavedra, H.I., Starovic, M., Trimboli, A., Yang, Y., Opav- promote apoptosis. J. Biol. Chem. 280, 21915–21923. ska, J., Wilson, P., Thompson, J.C., Ostrowski, M.C., et al. (2003). Extra- Sherr, C.J., and McCormick, F. (2002). The RB and p53 pathways in cancer. embryonic function of Rb is essential for embryonic development and viability. Cancer Cell 2, 103–112. Nature 421, 942–947.

194 Cancer Cell 15, 184–194, March 3, 2009 ª2009 Elsevier Inc.